/*
- Copyright (C) 2012 Carl Hetherington <cth@carlh.net>
+ Copyright (C) 2012-2021 Carl Hetherington <cth@carlh.net>
- This program is free software; you can redistribute it and/or modify
+ This file is part of DCP-o-matic.
+
+ DCP-o-matic is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
- This program is distributed in the hope that it will be useful,
+ DCP-o-matic is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
- along with this program; if not, write to the Free Software
- Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ along with DCP-o-matic. If not, see <http://www.gnu.org/licenses/>.
*/
+
#include "audio_decoder.h"
#include "audio_buffers.h"
-#include "exceptions.h"
+#include "audio_content.h"
+#include "dcpomatic_log.h"
#include "log.h"
+#include "resampler.h"
+#include "compose.hpp"
+#include <iostream>
#include "i18n.h"
-using std::stringstream;
-using std::list;
-using std::pair;
+
+using std::cout;
+using std::shared_ptr;
+using std::make_shared;
using boost::optional;
-using boost::shared_ptr;
+using namespace dcpomatic;
+
-AudioDecoder::AudioDecoder (shared_ptr<const Film> f, shared_ptr<const AudioContent> c)
- : Decoder (f)
- , _next_audio (0)
- , _audio_content (c)
+AudioDecoder::AudioDecoder (Decoder* parent, shared_ptr<const AudioContent> content, bool fast)
+ : DecoderPart (parent)
+ , _content (content)
+ , _fast (fast)
{
- if (_audio_content->content_audio_frame_rate() != _audio_content->output_audio_frame_rate()) {
+ /* Set up _positions so that we have one for each stream */
+ for (auto i: content->streams ()) {
+ _positions[i] = 0;
+ }
+}
- shared_ptr<const Film> film = _film.lock ();
- assert (film);
- stringstream s;
- s << String::compose (
- "Will resample audio from %1 to %2",
- _audio_content->content_audio_frame_rate(), _audio_content->output_audio_frame_rate()
- );
-
- film->log()->log (s.str ());
-
- /* We will be using planar float data when we call the
- resampler. As far as I can see, the audio channel
- layout is not necessary for our purposes; it seems
- only to be used get the number of channels and
- decide if rematrixing is needed. It won't be, since
- input and output layouts are the same.
- */
-
- _swr_context = swr_alloc_set_opts (
- 0,
- av_get_default_channel_layout (MAX_AUDIO_CHANNELS),
- AV_SAMPLE_FMT_FLTP,
- _audio_content->output_audio_frame_rate(),
- av_get_default_channel_layout (MAX_AUDIO_CHANNELS),
- AV_SAMPLE_FMT_FLTP,
- _audio_content->content_audio_frame_rate(),
- 0, 0
+/** @param time_already_delayed true if the delay should not be added to time */
+void
+AudioDecoder::emit(shared_ptr<const Film> film, AudioStreamPtr stream, shared_ptr<const AudioBuffers> data, ContentTime time, bool flushing)
+{
+ if (ignore ()) {
+ return;
+ }
+
+ int const resampled_rate = _content->resampled_frame_rate(film);
+ if (!flushing) {
+ time += ContentTime::from_seconds (_content->delay() / 1000.0);
+ }
+
+ /* Amount of error we will tolerate on audio timestamps; see comment below.
+ * We'll use 1 24fps video frame as this seems to be roughly how ffplay does it.
+ */
+ Frame const slack_frames = resampled_rate / 24;
+
+ /* first_since_seek is set to true if this is the first data we have
+ received since initialisation or seek. We'll set the position based
+ on the ContentTime that was given. After this first time we just
+ count samples unless the timestamp is more than slack_frames away
+ from where we think it should be. This is because ContentTimes seem
+ to be slightly unreliable from FFmpegDecoder (i.e. not sample
+ accurate), but we still need to obey them sometimes otherwise we get
+ sync problems such as #1833.
+ */
+
+ auto const first_since_seek = _positions[stream] == 0;
+ auto const need_reset = !first_since_seek && (std::abs(_positions[stream] - time.frames_round(resampled_rate)) > slack_frames);
+
+ if (need_reset) {
+ LOG_GENERAL (
+ "Reset audio position: was %1, new data at %2, slack: %3 frames",
+ _positions[stream],
+ time.frames_round(resampled_rate),
+ std::abs(_positions[stream] - time.frames_round(resampled_rate))
);
-
- swr_init (_swr_context);
+ }
+
+ if (first_since_seek || need_reset) {
+ _positions[stream] = time.frames_round (resampled_rate);
+ }
+
+ if (first_since_seek && _content->delay() > 0) {
+ silence (stream, _content->delay());
+ }
+
+ shared_ptr<Resampler> resampler;
+ auto i = _resamplers.find(stream);
+ if (i != _resamplers.end()) {
+ resampler = i->second;
} else {
- _swr_context = 0;
+ if (stream->frame_rate() != resampled_rate) {
+ LOG_GENERAL (
+ "Creating new resampler from %1 to %2 with %3 channels",
+ stream->frame_rate(),
+ resampled_rate,
+ stream->channels()
+ );
+
+ resampler = make_shared<Resampler>(stream->frame_rate(), resampled_rate, stream->channels());
+ if (_fast) {
+ resampler->set_fast ();
+ }
+ _resamplers[stream] = resampler;
+ }
}
-}
-AudioDecoder::~AudioDecoder ()
-{
- if (_swr_context) {
- swr_free (&_swr_context);
+ if (resampler && !flushing) {
+ /* It can be the the data here has a different number of channels than the stream
+ * it comes from (e.g. the files decoded by FFmpegDecoder sometimes have a random
+ * frame, often at the end, with more channels). Insert silence or discard channels
+ * here.
+ */
+ if (resampler->channels() != data->channels()) {
+ LOG_WARNING("Received audio data with an unexpected channel count of %1 instead of %2", data->channels(), resampler->channels());
+ auto data_copy = data->clone();
+ data_copy->set_channels(resampler->channels());
+ data = resampler->run(data_copy);
+ } else {
+ data = resampler->run(data);
+ }
+
+ if (data->frames() == 0) {
+ return;
+ }
}
+
+ Data(stream, ContentAudio (data, _positions[stream]));
+ _positions[stream] += data->frames();
}
-
-#if 0
-void
-AudioDecoder::process_end ()
+
+/** @return Time just after the last thing that was emitted from a given stream */
+ContentTime
+AudioDecoder::stream_position (shared_ptr<const Film> film, AudioStreamPtr stream) const
{
- if (_swr_context) {
-
- shared_ptr<const Film> film = _film.lock ();
- assert (film);
-
- shared_ptr<AudioBuffers> out (new AudioBuffers (film->audio_mapping().dcp_channels(), 256));
-
- while (1) {
- int const frames = swr_convert (_swr_context, (uint8_t **) out->data(), 256, 0, 0);
-
- if (frames < 0) {
- throw EncodeError (_("could not run sample-rate converter"));
- }
+ auto i = _positions.find (stream);
+ DCPOMATIC_ASSERT (i != _positions.end ());
+ return ContentTime::from_frames (i->second, _content->resampled_frame_rate(film));
+}
- if (frames == 0) {
- break;
- }
- out->set_frames (frames);
- _writer->write (out);
+boost::optional<ContentTime>
+AudioDecoder::position (shared_ptr<const Film> film) const
+{
+ optional<ContentTime> p;
+ for (auto i: _positions) {
+ auto const ct = stream_position (film, i.first);
+ if (!p || ct < *p) {
+ p = ct;
}
-
}
+
+ return p;
}
-#endif
+
void
-AudioDecoder::audio (shared_ptr<const AudioBuffers> data, Time time)
+AudioDecoder::seek ()
{
- /* Maybe resample */
- if (_swr_context) {
+ for (auto i: _resamplers) {
+ i.second->flush ();
+ i.second->reset ();
+ }
- /* Compute the resampled frames count and add 32 for luck */
- int const max_resampled_frames = ceil (
- (int64_t) data->frames() * _audio_content->output_audio_frame_rate() / _audio_content->content_audio_frame_rate()
- ) + 32;
+ for (auto& i: _positions) {
+ i.second = 0;
+ }
+}
- shared_ptr<AudioBuffers> resampled (new AudioBuffers (data->channels(), max_resampled_frames));
- /* Resample audio */
- int const resampled_frames = swr_convert (
- _swr_context, (uint8_t **) resampled->data(), max_resampled_frames, (uint8_t const **) data->data(), data->frames()
- );
-
- if (resampled_frames < 0) {
- throw EncodeError (_("could not run sample-rate converter"));
+void
+AudioDecoder::flush ()
+{
+ for (auto const& i: _resamplers) {
+ auto ro = i.second->flush ();
+ if (ro->frames() > 0) {
+ Data (i.first, ContentAudio (ro, _positions[i.first]));
+ _positions[i.first] += ro->frames();
}
-
- resampled->set_frames (resampled_frames);
-
- /* And point our variables at the resampled audio */
- data = resampled;
}
- shared_ptr<const Film> film = _film.lock ();
- assert (film);
-
- /* Remap channels */
- shared_ptr<AudioBuffers> dcp_mapped (new AudioBuffers (film->dcp_audio_channels(), data->frames()));
- dcp_mapped->make_silent ();
- list<pair<int, libdcp::Channel> > map = _audio_content->audio_mapping().content_to_dcp ();
- for (list<pair<int, libdcp::Channel> >::iterator i = map.begin(); i != map.end(); ++i) {
- dcp_mapped->accumulate_channel (data.get(), i->first, i->second);
+ if (_content->delay() < 0) {
+ /* Finish off with the gap caused by the delay */
+ for (auto stream: _content->streams()) {
+ silence (stream, -_content->delay());
+ }
}
-
- Audio (dcp_mapped, time);
- _next_audio = time + film->audio_frames_to_time (data->frames());
}
-
+
+void
+AudioDecoder::silence (AudioStreamPtr stream, int milliseconds)
+{
+ int const samples = ContentTime::from_seconds(milliseconds / 1000.0).frames_round(stream->frame_rate());
+ auto silence = make_shared<AudioBuffers>(stream->channels(), samples);
+ silence->make_silent ();
+ Data (stream, ContentAudio(silence, _positions[stream]));
+}